'You Are An Idiot!' - How Conversational Agent Communication Patterns Influence Frustration And Harassment

Conversational Agents (CA) in the form of digital assistants on smartphones, chatbots on social media, or physical embodied systems are an increasingly often applied new form of user interfaces for digital systems. The human-like design of CAs (e.g., having names, greeting users, and using self-references) leads to users subconsciously reacting to them as they were interacting with a human. In recent research, it has been shown that this social component of interacting with a CA leads to various benefits, such as increased service satisfaction, enjoyment, and trust. However, numerous CAs were discontinued because of inadequate responses to user requests or only making errors because of the limited functionalities and knowledge of a CA, which can lead to frustration. Therefore, investigating the causes of frustration and other related emotions and reactions highly relevant. Against this background, this study investigates via an online experiment with 169 participants how different communication patterns influence user's perception, frustration, and harassment behavior of an error producing CA.